Phenotypic analysis
The kernels of BL31 were normal and plump, whereas the Dek rate of BL33 is 100% in all environments (Additional file 1). The Dek rates in F2 and F2:3 populations were highly correlated among three environments with correlation coefficients ranging from 0.70 to 0.92 (Table 1). Variance analyses showed that both genotypes (P < 0.001) and environments (P < 0.05) had significant effects on the target trait (Table 2). The genotype × environment interactions did not reach a statistical significance (P > 0.05) (Table 2) and the broad-sense heritability of Dek rate in F2:3 across two environments was 0.95, denoting that the Dek rate was mainly determined by genotypes. The Dek rates in the populations showed continuous distribution (Fig. 1, a-d), indicating that Dek is controlled by multiple genetic loci. Additionally, the frequency graph underpinning the phenotypes is abnormal distribution with the “valley - peak” pattern, suggesting the presence of major QTL controlling Dek in the populations [16].
Chromosomal locations of Dek loci and construction of linkage maps
Theoretically, the more polymorphic markers between isogenic lines one chromosome contains, the higher probability that it harbors QTL for target traits. To rapidly identify the chromosomal locations of Dek QTL, we genotyped the isogenic lines, BL31 and BL33, with contrasting grain phenotypes using the Wheat660K SNP array. Ten chromosomes, 1A, 1B, 2A, 3A, 3B, 4A, 4B, 5A, 6B and 7B, were considered to be potential target chromosomes with Dek QTL based on the frequency distributions of polymorphic SNPs (Additional file 2). To further verify the chromosomal location of Dek QTL, SSR markers from the above target chromosomes were selected to construct genetic map. Totally 783 SSR markers on these chromosomes were selected to construct linkage map for QTL analysis (Additional file 3) [17-24]. Of them, 118 SSR markers displayed polymorphisms between two parents, while only 35 SSRs in chromosomes 3B (13), 4A (19) and 5A (3), showed consistent polymorphisms between two groups of lines with contrasting phenotypes, i.e. 40 defective kernel and 40 normal kernel lines, respectively. Subsequently, the polymorphic SSR markers were used to genotype the entire F2 population, and two linkage groups were constructed in chromosomes 3B and 4A, including 6 and 9 SSR markers, respectively (Fig. 2).
QTL mapping
Based on the resultant linkage groups above and Dek rates in the genetic populations, we performed QTL mapping for Dek traits; three QTL were detected on chromosomes 3B and 4A, designated as QDek.caas-3BS.1, QDek.caas-3BS.2 and QDek.caas-4AL, respectively (Table 3; Fig. 2). QDek.caas-3BS.1 was flanked by Xbarc133 and Xcfd79 spanning a genetic interval of 6.77 cM; QDek.caas-3BS.2 was located between Xcfd79 and Xwmc808 in an interval of 35.03 cM; QDek.caas-4AL was mapped between Xwmc500 and Xgpw3079 in a genetic interval of 15.17 cM (Fig. 2). Based on the preliminary mapping above, the physical regions of QDek.caas-3BS.1, QDek.caas-3BS.2 and QDek.caas-4AL can be defined by their flanking markers according to IWGSC reference genome RefSeq v1.0 (http://plants.ensembl.org/Triticum_aestivum/Info/Index) [15]. The physical intervals of QDek.caas-3BS.1, QDek.caas-3BS.2 and QDek.caas-4AL are 7.60 - 18.84 Mb, 18.84 - 32.75 Mb in chromosomes 3BS and 704.34 - 710.05 Mb in chromosome 4AL. Genetic analyses showed that QDek.caas-3BS.1, QDek.caas-3BS.2 and QDek.caas-4AL could explain 14.78 to 18.17%, 16.61 to 21.83% and 19.08 to 28.19% of the phenotypic variances, respectively (Table 3). Each of these QTL determined more than 10% phenotypic variation of Dek in each environment, indicating that they are major QTL, consistent with the “peak - valley” distribution pattern of Dek rate in genetic populations. Additionally, all the alleles governing the formation of Dek came from BL33.
Marker enrichment in the target regions of QTL
To further narrow down the genetic intervals of target QTL, the polymorphic SNPs identified from the Wheat660K chips were selected according to the above physical regions defined by QTL preliminary mapping. Totally 60 polymorphic SNPs were used to develop site-specific CAPS markers and five of these, AX-109027972, AX-109516011, AX-110042240, AX-109301653 and AX-108996126, were successfully mapped in the linkage groups (Additional file 4; Additional file 5; Additional file 6). Subsequently, QDek.caas-3BS.2 and QDek.caas-4AL were narrowed to the genetic intervals of 14.01 cM and 13.18 cM, respectively (Table 3; Fig. 2). Accordingly, the physical intervals of QDek.caas-3BS.2 and QDek.caas-4AL are reduced to 1.16 Mb and 1.13 Mb, respectively (Table 3). Although the polymorphic loci AX-108996126 and AX-110042240 did not further diminish the genetic interval of QDek.caas-3BS.1 and QDek.caas-4AL, they saturated the target regions (Fig. 2).
Dek trait displays abnormal development of grain and thus it is usually considered to affect grain weight. To verify this case, we analyzed the effect of the resultant Dek QTL on thousand grain weight (TGW). Compared with BL31, the TGW of BL33 significantly decreased 24.12 - 28.81% (Additional file 7). QTL analyses based on the linkage maps above showed that only one QTL for TGW, designated as QTGW.caas-3BS, was detected between Xcfd79 and AX-109027972, explaining 19.18 to 23.94% of phenotypic variances (Additional file 8). QTGW.caas-3BS overlapped QDek.caas-3BS.2 on the linkage map, suggesting that the Dek QTL QDek.caas-3BS.2 may affect TGW. No significant effect on TGW was detected in the genetic intervals of the other two QTL, QDek.caas-3BS.1 and QDek.caas-4AL.
Predication of candidate genes for the Dek QTL
The cloned dek genes in maize are involved in the grain growth and development, especially the synthesis and storage of starches and/or proteins in the endosperm [4-7]. Our biochemical analysis also showed that the mature grain starch contents of BL33 were significantly lower than those of BL31 (Additional file 9). In contrast, BL33 had higher sucrose contents and slower sucrose reduction across the filling stages than BL31 (Additional file 10). Accordingly, we assumed that the causal genes for the three QTL are involved in carbohydrate metabolism. In this study, three major QTL was mapped into the genetic intervals of less than 15 cM. The physical region of QDek.caas-3BS.1 is 11.24 Mb, whereas those of QDek.caas-3BS.2 and QDeg.caas-4AL are only 1.16 and 1.13 Mb, respectively, according to IWGSC RefSeq v1.0 (Table 3). Totally 233 genes were present in the physical intervals of the three QTL (Additional file 11). We found nine genes (Gene ID: TraesCS3B01G025200, TraesCS3B01G028100, TraesCS3B01G028200, TraesCS3B01G028300, TraesCS3B01G028500, TraesCS3B01G028700, TraesCS3B01G039100, TraesCS3B01G039800, and TraesCS4A01G446700) involving carbohydrate metabolism or grain development in or near the target regions of the Dek QTLs (Additional file 11).
To further identify candidate genes for the Dek QTL, the spatio-temporal expression patterns of the nine genes were analyzed using the WheatExp (https://wheat.pw.usda.gov/WheatExp/). Of these, four genes, TraesCS3B01G025200, TraesCS3B01G028700, TraesCS3B01G039800, and TraesCS4A01G446700, had an observable expression during grain development (Additional file 12) [25]. Based on functional annotation in IWGSC RefSeq v1.0, the four genes encoded fructose-bisphosphate aldolase (Fba), β-fructofuranosidase, abscisic acid-deficient 4 (ABA4), and sucrose synthase (Sus), respectively, tentatively designated as TaFba-3B, TaBff-3B, TaABA4-3B, and TaSus-4A. Their expression patterns were confirmed in the immature grains harvested at three different stages (5, 15, and 25 DAF) using qPCR (Additional file 13). The results showed that all of the above genes expressed in grain and had the lowest expression level at 5 DAF. TaBff-3B and TaSus-4A was mainly detected at 15 DAF, while TaFba-3B and TaABA4-3B had the highest expression activity in 25-DAF grains.